【摘要】：The recurving cluster of tropical cyclones( TCs) landing at China in spring has the characteristics of high frequency,strong intensity,severe influence and long lifespan,a better understanding of tropical cyclone tracks and their underlying mechanisms and more accurate prediction are thus of great value for the prevention and mitigation of TC-related disasters. Based on the best track dataset of typhoon during 1951-2018 from Shanghai Typhoon Institute( STI) of Chinese Meteorological Administration,the tracks of tropical cyclones landfalling over the Chinese coast during 1951-2018 are grouped into three clusters through finite mixture model. The ENSO index,PDO index and the 74 circulation indices of the National Climate Center are analyzed by the classification and regression tree( CART) algorithm for the first TC cluster( the northward-moving cluster). The prediction model based on the results of CART for spring track lifespan is built by a random sampling of the data during the 50 years( about 75%) as a model of the training set,and the training accuracy is 82. 46%. The 17-year( about 25%) remaining data are used for testing,with a prediction accuracy of 75%. This study suggests that the finite mixture model algorithm produces a better classification of the northward-going-cluster tracks of TC landfalling over China. In addition,the CART algorithm,which is used for the classification of track lifespan prediction model,not only shows high accuracy,but also the results are easily explained and understood. It provides a novel idea for forecasting the lifespan of tropical cyclones landing at China.